Characterization of the tandem CWCH2 sequence motif: a hallmark of inter-zinc finger interactions

Background The C2H2 zinc finger (ZF) domain is widely conserved among eukaryotic proteins. In Zic/Gli/Zap1 C2H2 ZF proteins, the two N-terminal ZFs form a single structural unit by sharing a hydrophobic core. This structural unit defines a new motif comprised of two tryptophan side chains at the center of the hydrophobic core. Because each tryptophan residue is located between the two cysteine residues of the C2H2 motif, we have named this structure the tandem CWCH2 (tCWCH2) motif. Results Here, we characterized 587 tCWCH2-containing genes using data derived from public databases. We categorized genes into 11 classes including Zic/Gli/Glis, Arid2/Rsc9, PacC, Mizf, Aebp2, Zap1/ZafA, Fungl, Zfp106, Twincl, Clr1, and Fungl-4ZF, based on sequence similarity, domain organization, and functional similarities. tCWCH2 motifs are mostly found in organisms belonging to the Opisthokonta (metazoa, fungi, and choanoflagellates) and Amoebozoa (amoeba, Dictyostelium discoideum). By comparison, the C2H2 ZF motif is distributed widely among the eukaryotes. The structure and organization of the tCWCH2 motif, its phylogenetic distribution, and molecular phylogenetic analysis suggest that prototypical tCWCH2 genes existed in the Opisthokonta ancestor. Within-group or between-group comparisons of the tCWCH2 amino acid sequence identified three additional sequence features (site-specific amino acid frequencies, longer linker sequence between two C2H2 ZFs, and frequent extra-sequences within C2H2 ZF motifs). Conclusion These features suggest that the tCWCH2 motif is a specialized motif involved in inter-zinc finger interactions.


Background
Zinc finger (ZF) domains (ZFDs) are found in a large number of eukaryotic proteins [1][2][3]. A single ZF normally forms a globular structure that is stabilized by binding a zinc ion. Many classes of ZFs have been described in public databases (Pfam, http://pfam.sanger. ac.uk/; Prosite, http://au.expasy.org/prosite/; Interpro, http://www.ebi.ac.uk/interpro/; SMART, http://smart. embl-heidelberg.de/), with Cys-Xn-Cys-Xn-His-Xn-His (C2H2) being one of the most common. Most C2H2 ZF proteins contain tandem arrays of the C2H2 motif, which are located at specific intervals to form a functional domain. ZFDs were originally identified as the DNA-binding domain of transcription factor IIIA (TFIIIa) and other transcription factors, but accumulating evidence suggests that ZFDs also bind RNA and proteins [4][5][6]. Since ZFDs are essential for many biological processes, their structure-function relationships have been well studied. For the DNA-binding ZFs, the structures of ZFD-DNA complexes have been elucidated for GLI [7], TFIIIA [8], zif268 [9], YY1 [10], and WT1 [11]. These findings have enabled the development of synthetic ZF proteins as versatile molecular tools [12]. Despite their important and widespread roles, the structural basis of ZFD-protein interactions is less understood than those of ZFD-DNA interactions.
The identification of the critical structural features of protein-binding ZFDs remains elusive [5] but some clues are available for a group of ZFs that mediate protein-to-protein interactions [4][5][6]. Previously, we studied the intra-molecular protein-to-protein interaction between two adjacent C2H2 ZFs [13]. The two N-terminal ZFs (ZF1 and ZF2) of human ZIC3, which has a ZFD composed of five ZFs (ZF1-ZF5), form a single structural unit through a common hydrophobic core. This ZF-connecting hydrophobic core is associated with two tryptophan residues, each of which is located between the zinc-binding cysteine residues in ZF1 and ZF2. Mutation of the tryptophan in ZF1 (W255G) perturbs the subcellular localization and function of the protein, and is associated with a pathological and congenital heart malformation [13,14]. The importance of these tryptophan residues is further supported by their conservation among more than 40 Zic proteins identified in a wide range of eumetazoan species [15].
Previous studies have revealed that human GLI1 (PDBID: 2GLI) and yeast Zap1 (PDBID: 1ZW8) also possess ZFs with two tryptophans in the corresponding region and that these tryptophans are located in the hydrophobic core formed by two adjacent ZFs [16]. These data raise the possibility that domains containing the consensus sequence "Cys-X-Trp-Xn-Cys-Xn-His-Xn-His-Xn-Cys-X-Trp-Xn-Cys-Xn-His-Xn-His", which we have named the tandem CWCH2 (tCWCH2) motif, are involved in the interaction between the two adjacent ZFs. Since our knowledge of the tCWCH2 motif is limited to a small group of proteins, the biological significance of the tCWCH2 structure is unclear.
In the present study, we performed a computer-based analysis of the tCWCH2 motif using sequence data derived from public databases. We classified the tCWCH2-containing sequences into gene classes and then determined their conservation status in each protein family and their phylogenic distribution. We also identified sequence features unique to tCWCH2-containing ZFDs. The significance of the tCWCH2 motif is discussed in terms of its possible structural and functional roles.

Three-dimensional structure of known tCWCH2 motifs
To investigate the positions of the two key tryptophan residues in different tCWCH2 motifs, we performed a structural alignment to compare the three-dimensional (3D) structures of the tCWCH2 motifs from ZIC3 (PDBID: 2RPC), GLI1 (PDBID: 2GLI) and Zap1 (PDBID: 1ZW8) ZFDs ( Figure 1A) with that of two C2H2 ZFs. Our data show that all of these C2H2 ZF motifs form globular structures composed of two anti-parallel β sheets and an α helix (ββα). The two tryptophan residues between the two zinc-chelating cysteine residues ( Figure 1B) were localized onto the anti-parallel β sheet where they were juxtaposed to each other. The relative positions of the two tryptophans were similar among the three tCWCH2 structures, as was each zinc-chelating residue. The tryptophan side chains were close to the zinc-chelating histidine in the same ZF, and the hydrophobic residues in the same and the opposite ZF (within 5Å, Additional file 1). This result suggested that the tryptophans play a role in both the stabilization of their own ZFs and the formation of a hydrophobic core between the two adjacent ZFs. The conserved positioning of the tCWCH2 motif-forming residue in evolutionarily distant sources led us to investigate its phylogenetic distribution and structural features.

Database search for tCWCH2-containing sequences
We performed a comprehensive search of tCWCH2 motif-containing genes in current databases. We performed an initial search of the non-redundant NCBI protein database with PHI-BLAST using the tCWCH2 pattern deduced from the ZIC comparison and human ZIC3 (NP_003404) as the query sequence. This search yielded a set of 709 amino acid sequences that were tentatively classified into 11 gene classes according to their annotations and the pilot phylogenetic tree analyses (see Methods, Figure 2). For the fine analysis of the tCWCH2-containing sequences, we performed TBLASTN and PBLAST searches of the non-redundant sequence collection of the NCBI database using the representative sequences of the initial 11 gene classes and 3 specific genes (two Dictyostelium genes and one Monosiga gene) as the key sequences. In order to examine the distribution of the tCWCH2 sequence motif in a phylogenetic tree of eukaryotes, TBLASTN searches were also performed on a whole-genome shotgun sequence database of 24 organisms. The sequences obtained were manually checked to remove any redundant sequences. The final tCWCH2 collection contained 587 sequences that represented the non-redundant tCWCH2-containing sequences from a wide range of organisms (Additional file 2). Because the Mizf and Zap1/ZafA family genes and one Dictyostelium gene contain two independent tCWCH2 motifs, our collection included a total of 637 independent sequence regions for tCWCH2. At this point, we re-evaluated the initial sequence pattern for PHI-BLAST because it was based solely on the sequence compilation of the Zic family genes. The initial PHI-BLAST pattern recovered 90% (572/637) of the collected sequences (Additional file 3). Among the 65 nonmatching sequences, 20 sequences showed deviation in the C, W, and H residues of the tCWCH2 motif and 50 sequences showed derangement of the sequence length between the two CWCH2 motifs. This mismatch appeared to be distributed randomly and suggested that optimization of the PHI-BLAST pattern was unlikely to be informative. Therefore, we optimized the lengths of the intervening sequences. PHI-BLAST searches using the original pattern "xCxWx(2,35)Cx(5,15)Hx(2,5)Hx (5,25)CxWx(1,3)Cx(5,17)Hx(2,5)Hx" with a tCWCH2 consensus sequence (see Methods) yielded 911 sequences, of which 459 sequences were included in our non-redundant tCWCH2 collection. This number corresponded to 83% of the tCWCH2 sequences in the NCBI protein database (total = 551). A looser pattern, "xCxWx(1,35)Cx(5,15)Hx(2,5)Hx(5,42)CxWx(1,6)Cx (5,31)Hx(2,5)Hx" yielded 968 sequences that contains 469 tCWCH2 sequences (85% of the total) (Additional file 3). The sequences identified by the looser pattern included Zfp407 sequences that contain a pair of CWCH2 ZFs separated by another two C2H2 ZFs. These data suggest that decreasing the stringency of the search (increasing the intervening arbitrary residue number) impairs the fidelity of the search. Therefore, we consider that the original PHI-BLAST pattern is more useful for the tCWCH2 survey.
Classification and domain structure of tCWCH2 sequence-containing proteins We classified the non-redundant tCWCH2-containing sequences based on the available flanking amino acid sequences of the tCWCH2 domain ( Figure 2). Each gene contained one or two tCWCH2 motifs among a total of 2-9 C2H2 ZF motifs. Proteins containing only one tCWCH2 domain suggest that a single tCWCH2 constitutes the minimal ZFD component. In the ZFDs composed of both tCWCH2 and non-tCWCH2 C2H2, tCWCH2s were always placed at the N-terminal end of the clustered ZFDs.

Arid2/Rsc9
Arid2/Rsc9 genes are known to encode components of an ATP-dependent chromatin remodeling complex  [ [27][28][29]. The ARID (AT-rich interaction domain) and tCWCH2 domains are conserved in Arid2 and Rsc9 (Additional file 4). Another conserved sequence motif was found in the C-terminal flanking region of the tCWCH2 motif. This sequence motif is summarized as "IxL (S/T) AxL (I/V) L (K/R) N (I/L) × (K/R)" and we named this motif the LIL domain. PHI-BLAST searches for LIL domains suggested that this domain was distributed only in the Arid2/Rsc9 proteins (data not shown).

PacC
PacC mediates gene expression regulation by ambient pH in filamentous fungi and yeasts [30,31]. It contains an N-terminal DNA binding domain with three ZFs. Other ZF domains, named A, B, and C, are involved in responses to pH change [32,33].

Mizf
Mizf was originally reported as a seven ZF-containing protein [34][35][36]. We identified two additional C2H2 ZFs in the gene sequence encoding this protein (ZF2 and ZF8 in Additional file 5). The two tCWCH2 motifs in the first four ZFs (ZF1-ZF4) were separated by longer intervening sequences (>35 aa) than most ZFs.

Fungus-Gli-like (Fungl)
The amino acid sequences encoded by this group of genes have been known as "fungus Gli like". Here, we have named these sequences Fungl (Fungus-Gli-like) (Additional file 7, data not shown).
Tandem-CWCH2-protein-C-terminal (Twincl) The sequences are composed of 468 to 1458 amino acid residues (data not shown). On the basis of these structural features (Figure 2, Additional file 8), we named this gene family Twincl (Tandem CWCH2 protein C-terminal).

Clr1
The ZFs among members of the Clr1 family [56] were highly divergent and other sequences were not conserved (Additional file 9, data not shown). Therefore, we omitted this gene family from the consensus sequence and phylogenetic tree analyses.
This gene group includes annotations from Rhizopus oryzae and Batrachochytrium dendrobatidis representing the fungal phyla Mucoromycotina and Chytridiomycota, respectively. The proteins encoded by the Fungl-4ZF gene family contain four ZFs that are weakly similar to those in Gli and Fungl (Additional file 7). Based on the ZF motif organization, we tentatively grouped them separately from Fungl. However, based on their sequence similarity and phylogenetic classification it is possible that the Fungl-4ZF and Fungl gene groups are derived from a common ancestral gene [57,58].

Dictyostelium discoideum tCWCH2
The social amoeba Dictyostelium discoideum belongs to the Amoebozoa supergroup. There are two tCWCH2containing genes in this gene class, DdHp1 (Dictyostelium discoideum hypothetical protein) and DdHp2. The ZFD alignment of these genes is shown in Additional file 10.

Monosiga brevicollis tCWCH2
Since Monosiga brevicollis is the only species among the choanoflagellates in which the genome has been fully sequenced, the presence of one tCWCH2 sequence (Additional file 10) may be more functionally meaningful than the singleton sequences in metazoa and fungi. The existence of an RFX domain (position 376-423) raised the possibility that the tCWCH2 sequence in Monosiga brevicollis and Arid2/Rsc9 is derived from a common ancestral gene.

Unclassified sequences
Six sequences were not classified into the above 13 categories (Additional file 2). Four out of the six unclassified sequences contained weak similarities to PacC, Fungl, Fungl-4ZF, and Mizf. The remaining two sequences from fungi did not show any significant homology to the other sequences.

Isolated tCWCH2 in metazoan gene families
We found rare occurrences of the tCWCH2 motif in some metazoan gene families, including three tCWCH2containing sequences in the Sp8 family (n = 37), one in the TRP2 family (n = 8), one in the Krüppel-like factor 5 family (n = 14), and one in the pbrm-1 family (n = 21) (Additional file 2). We did not include these genes in the following analysis because their significance was not clear.

tCWCH2 sequences in plant databases
Two plant tCWCH2-containing sequences were found. One sequence (XP_001771543) was the REF6 homolog of a moss (Bryophyta, Physcomitrella patens) (Additional file 10). Although REF6 is conserved in many plants, the CWCH2 was not found in other plant species including Arabidopsis thaliana, Oryza sativa, and Vitis vinifera (Additional file 10). The other sequence, AK110182, was from Oryza sativa but is highly homologous to PacC. This sequence was otherwise only detected in fungi and therefore might represent contamination of Oryza sativa with fungal material.
In the course of the classification procedure, we extracted the sequences belonging to each gene family irrespective of the presence of tCWCH2 motifs identified by BLAST search using representative entire amino acid sequences. Within this group, we examined the extent of conservation of the cysteine, tryptophan, and histidine residues contributing to the tCWCH2 motifs. The analysis indicated that the tryptophans in CWCH2 motif-constituting residues were generally conserved (87%-100%) ( Table 1).

Similarity among tCWCH2 gene classes
We performed molecular phylogenetic analyses to identify similarities among the tCWCH2 gene classes. For this purpose, we selected flanking ZFs containing tCWCH2 (tCWCH2+1ZF) because the tCWCH2 itself yielded very little information on sequence similarities, presumably because it is short. In addition, genes widely distributed among fungi or metazoa were used for this analysis because our preliminary analysis suggested that genes with a narrow phylogenetic distribution (Twincl, Clr1, Fungl-4ZF, Dictyostelium tCWCH2) had strongly divergent sequences that may confound the phylogenetic tree analyses [59]. Using the tCWCH2+1ZF (3ZFs) amino acid sequences, we constructed phylogenetic trees by the Bayesian inference (BI), maximal likelihood (ML), and neighbor joining (NJ) methods. The 89 (80 tCWCH2 + 9 TFIIIA) sequences, representing a wide phylogenic range, were subjected to this analysis using the DNA-binding classical C2H2 ZFs of TFIIIA as outgroup sequences ( Figure 3; sequence alignment in Additional file 11; detailed tree in BI, ML, NJ in Additional files 12, 13 and 14). Each class of tCWCH2-containing genes was strapped with high bootstrap values in the BI, ML, and NJ methods, supporting the validity of the above classification. Phylogenetic tree analyses revealed the relationships among the tCWCH2 gene families and revealed a novel relationship among the metazoan Zic, Gli, and Glis gene families. These three gene families were grouped together, whereas Gli and Glis formed a sub-group in the branch of the Zic/Gli/Glis group. It is known that vertebrate Zic, Gli, and Glis are composed of six, three, and three paralogs, respectively. Interestingly, Glis2 and Glis3 were separately grouped with their insect homologs (DmGlis2 and DmGlis3 respectively) and sea anemone Nematostella homologs (NveGlis2 and NveGlis3 respectively; Additional files 12, 13 and 14), indicating that the two paralogs existed in the common ancestors of cnidarians and bilaterians. On the other hand, vertebrate Zic and Gli paralogs were not grouped with the invertebrate homologs (Additional files 12, 13 and 14), suggesting that vertebrate Zic and Gli paralogs were generated in vertebrate ancestors.
Fungl was determined to have sequence similarity with the Zic/Gli/Glis group. The Fungl group branch was always located closest to the Zic/Gli/Glis group in the BI, ML, and NJ trees, but statistical support was weak (BI, 93%). The other gene groups with more than two classes of tCWCH2 motif were not strongly supported (BI, <70%). However, Zap1/ZafA class and Aebp2 class proteins were grouped together by the BI and ML methods, and the PacC family was placed in the basal root of the BI and ML trees beside the other tCWCH2+ZF1 sequences from metazoa and fungi.
Distribution of the tCWCH2 sequences in the phylogenetic tree Based on the above analysis, we investigated the phylogeny of the tCWCH2 motif ( Figure 3). In global terms, tCWCH2 sequences were detected in all of the organisms we examined in the Opisthokonta supergroup, The percentages indicate conservation of the tryptophan residues in each gene class. In Mizf and Zap1/ZafA, there are two independent tCWCH2 motifs, for which ZF12 (ZF1-ZF2) indicates the N-terminally located one and ZF34 (ZF3-ZF4) indicates the C-terminally located one.
In the Amoebozoa supergroup, we found two tCWCH2containing genes in the social amoeba Dictyostelium discoideum, but not in Entamoeba histolytica. Apart from Opisthokonta and Amoebozoa, two sequences were found in the Plantae databases, but their meaning is not clear. Thus, it appears that the distribution of the tCWCH2 sequence motif is essentially limited to the Uniconta (a collective term for the Opisthokonta and Amoebozoa supergroups) in current sequence databases. We examined the distribution of each tCWCH2 motif class in the phylogenetic tree ( Figure 4). Arid2/Rsc9 class genes were most widely detected in both fungi and metazoa. On the other hand, tCWCH2 was not detected in the sequences from the order Saccharomycetales (NP_013579, Saccharomyces cerevisiae; XP_718578, Candida albicans; XP_506030, Yarrowia lipolytica). Zic/Gli/Glis, Aebp2, and Mizf were widely distributed in metazoan groups, but Zfp106 was detected only in Bilateralia except Ecdysozoa. PacC, Zap1/ZafA, and Fungl class genes were widely distributed among the fungi groups. Distribution of Twincl class genes was restricted to Aspergillus and its close relatives. Clr1 class sequences were found only in Schizosaccharomyces and Cryptococcus. Fungl-4ZF class genes were recovered only from the basal fungi (Rhizopus oryzae and Batrachochytrium dendrobatidis).

Additional sequence features in the tCWCH2 motifs
We generated tCWCH2 consensus sequences for each gene class as well as for all classes to identify any additional sequence features. The Prosite database PDOC00028 alignment was used for the reference consensus sequence of general C2H2. The extent of conservation is indicated by the size of letters, and the consensus sequences are aligned graphically ( Figure 5). Classical C2H2 ZFs are known to have a consensus sequence of "(F/Y)xCx2CxFx7Lx2Hx4H" [1]. The tCWCH2 motifs also contain conserved phenylalanine and leucine residues between the cysteine and histidine residues. Phenylalanine and leucine are hydrophobic residues that generally mediate a hydrophobic interaction within a single ZF (data not shown). These data indicate that the general structure of a single ZF is well conserved in the tCWCH2 motif ( Figure 6A) and that there are some tCWCH2-specific structures apart from the conserved tryptophan residues.
Firstly, a position-specific bias to hydrophobic residues was found in the C-terminal region adjacent to the first histidine residue in ZF1 and ZF2 ( 1, 2 in Figure 5, 6A). Valine, leucine, and isoleucine were frequently found at 1 position (34.1%, 22.0%, and 35.3%, respectively), and valine, leucine, isoleucine, and methionine were frequently found at 2 position (14.3%, 14.3%, 39.1%, and 24.2%, respectively) in contrast to those in a compilation of general C2H2 sequences (PDOC00028) (valine, 4.2%; leucine, 9.5%; isoleucine, 5.3%; methionine, 7.8%; Figure 6B). The difference in the summed frequency of each position was statistically significant ( 1, P < 1 × 10 -100 ; 2, P < 1 × 10 -100 in a χ 2 test). These positions were close to the tryptophan residues on the opposite side of the ZF as determined in the superimposed 3D structures of Zic, Gli, and Zap1 tCWCH2s ( Figure 6A). The biased residue frequency may reflect the spatial restriction of amino acid choices. Since the CWCH2 structure suggests the presence of four hydrophobic residues in a compact space, the residue paired with each tryptophan could not be a residue with a large side chain, such as phenylalanine or tryptophan.
Thirdly, extra sequences were inserted between the tryptophan and cysteine residues in tCWCH2. This type of insertion was limited to the sequences from the first ZF of Zic/Gli/Glis, Arid2/Rsc9, and Zap1/ZafA and the second ZF of Arid2/Rsc9 classes. The number of amino acid residues between the two cysteines of tCWCH2 varied from four to 34 and was mostly four residues in other classes [15]. This insertion forms an extra looping structure in ZIC3 (PDBID: 2RPC; Figure 1, 6A) [13].

Discussion
The tCWCH2 motif is a hallmark of inter-zinc finger interactions The tCWCH2 motif was originally proposed as an evolutionarily conserved sequence involved in the interaction between ZFs in human ZIC3 [13,15]. The presence  Sequence conservation among the classes of tCWCH2 sequence motifs. Amino acid sequences were aligned and consensus sequences were generated (see Methods). Mizf ZF1-2, Mizf ZF3-4, and Zap1/ZafA ZF1-2 and ZF3-4 are indicated separately. The PDOC00028 C2H2 consensus is shown at the bottom as a general C2H2 consensus sequence (indicated by the tandem repeat of the consensus sequence). The short insertion sequences listed below were initially located at the sites indicated by the colored arrowheads in the alignments described above, but were separated to allow more comprehensive analyses. These sequences represent the longer linker sequences and the insertion of extra sequences described in the Results section. To achieve this analysis we omitted the four sequences AAWT01013938 (Schmidtea mediterranea Aebp2), CAG05504 (Tetraodon nigroviridis Gli), XP_001602003 (Nasonia vitripennis Gli), and XP_785526 (Strongylocentrotus purpuratus Gli) because these sequences contain exceptionally divergent sequences that disrupt the alignments.
of the tCWCH2 motif in the Gli family, Glis family, Zap1, and PacC has been previously reported [13]. Superimposition analysis revealed a tightly conserved structure among ZIC3, GLI, and Zap1, supporting the importance of the tCWCH2 motif in structural terms. Strong conservation of the tCWCH2 tryptophan residues in each gene class suggests that there has been strong selective pressure for them during evolution.
Our analysis revealed additional structural features of tCWCH2 besides the conserved tryptophans, which may be explained by adaptation of its role in mediating the inter-ZF interaction ( Figure 8). Firstly, there is a preference for valine, leucine, or isoleucine in the 1 and 2 position. This can be regarded as an adaptive adjustment for forming hydrophobic cores of the appropriate size. Furthermore, hydrophobic interaction between 1 and 2 is likely because they are closely located within 4Å distance in the 3D structure models of ZIC3, GLI and Zap1 (Additional file 1, data not shown). This also can influence the amino acid preferences in 1 and 2 position. Secondly, elongation of the linker sequences may also reflect a structural adaptation to allow the motif to face the two globular C2H2 units. Linker length and sequence are important for the function of the DNA-binding C2H2 ZF, because they determine the distance of each ZF and the linker contacts to the DNA base [60][61][62]. However, optimization of the canonical linker length for DNA-binding may not be enough to yield opposed positioning of ZF units. The bending between the two ZFs may be larger than is dictated by the DNA binding state of the ZFs as revealed in the GLI-DNA complex structure [7]. Longer linker size seems to be a general feature of tCWCH2, but conservation of the linker sequences within each class varies notably. Linker conservation in PacC is particularly high, despite the widespread distribution of this gene in fungi, thus suggesting that the linker in PacC is under evolutionary constraint.
Another structural feature, the addition of extra sequences, is plausible if we assume the reduction of structural constraint as a consequence of specialization in the inter-ZF interaction. For example, the conservation of tCWCH2-forming ZF1 is lower than that of ZF2-5 in Zic proteins [15]. In the case of the GLI-DNA complex, the tCWCH2-forming ZF1 does not directly participate in DNA binding [12], and the sequence conservation of Gli family ZF1 is also lower than in ZF2-5 ( Figure 5 It is unclear if tCWCH2 is the only structure required for inter-ZF interactions. The C2H2 ZFD is capable of achieving protein-to-protein interactions in many cases [6]. Interestingly, in a previous study to design synthetic  Comparison of linker length between the tCWCH2 sequence motif and general C2H2 ZF. The bars indicate the mean lengths of the linker sequence between C2H2 motifs in the indicated groups. We defined the linker length as the number of amino acids between the last histidine of a ZF and the first cysteine of the C-terminally flanking ZF. The PDOC00028 (general C2H2) alignment contained 12326 sequences and we found 10025 linkers in it. The general C2H2 sequences was further divided into two groups based on the presence or absence of "TGE (K/R)P". The sequence numbers of each group: tCWCH2, 614; general C2H2, 10025; "TGE(K/P)", 4640; "non-TGE(K/R)P", 5385). Error bar, standard deviation. *, P < 1 × 10 -100 in Mann-Whitney U-test. ZFs utilizing in vitro-evolving protein-to-protein interactions [16], it was found that the hydrophobic residues in zif268 that interacted between the artificial proteins and the zif268 C2H2 ZF were similar to those involved in the GLI inter-ZF interaction. Superimposition revealed that the position corresponding to the tryptophan in tCWCH2 is occupied by a valine, suggesting that either intramolecular or intermolecular protein-toprotein interactions, mediated by hydrophobic residues, are generally utilized for C2H2 ZFs. However, the strong conservation of the tCWCH2 and accompanying structural features suggest that tCWCH2 is a highly specialized C2H2 ZF for inter-ZF interactions between two adjacent ZFs.

Evolution of the tCWCH2 motif
We hypothesize that the original tCWCH2 motifs were generated from classical C2H2 ZFs (Figure 8) that occur widely in eukaryotic organisms, including Plantae, Excavata, Chromalveolata, and Rhizaria. tCWCH2 motifs were found in the vicinity of C2H2 ZF in several tCWCH2 gene classes. It is proposed that the tCWCH2 sequence appeared in the common ancestor of Opisthokonta or Unikonta (Opisthokonta/Amoebozoa). The acquisition of the tryptophans and additional structural features of tCWCH2 may have occurred concurrently in the course of evolution. Whether the evolution of tCWCH2 occurred once or multiple times during Opisthokonta/Amoebozoa evolution remains unclear. However, its short sequence and some isolated distribution in the plant supergroup and metazoan gene families may favor the multiple-origins hypothesis. Although the lineage relationship among the tCWCH2 classes remains unclear, we can presume the presence of the following tCWCH2 ancestral genes: an Arid2/Rsc9 common ancestor, which may have existed before the diversification of metazoa and fungi; a Zic/Gli/Glis common ancestor, an Aebp2 ancestor, and a Mizf ancestor in the early metazoa; a Fungl ancestor, a PacC ancestor, a Fungl-4ZF ancestor, and a Zap1/ZafA ancestor in the early fungi; a Zfp106 ancestor in early bilaterians; a Twincl ancestor in the founder of Aspergillus and its close siblings; and a Clr1 ancestor in the founder of Dikarya, althogh most Dikarya may have lost the gene. Once the tCWCH2 structures were established, they may have been more strongly conserved in these classes.

Function and role of tCWCH2
There are several cases that directly indicate the functional importance of the tCWCH2 motif in living organisms. In the case of PacC, mutation of the tryptophan of tCWCH2 induces a loss of its DNA binding activity [31,32]. In case of human ZIC3, the tryptophan mutation is proposed to be associated with the occurrence of a familial congenital heart defect, and biochemical analysis revealed that transcriptional activity and protein stability are decreased in the mutant protein [13,14]. In yeast Zap1, interaction between the two ZF where the tCWCH2 motif is located is required for the high-affinity binding of the zinc ion [45]. Also, it has been shown that the inter-ZF interaction is influenced by the concentration of zinc ion [46], suggesting that the tCWCH2 motif is involved in zinc-sensing in yeast.
Apart from the instances referred to above, the functional significance of the tCWCH2 domain remains unknown. We propose that tCWCH2 may play roles other than the direct DNA binding that is often exerted by classical C2H2, because the linker lengths of tCWCH2 are more highly divergent than those of the canonical C2H2 ZFs, in which linker length is a critical parameter for DNA-C2H2 ZF interactions [60][61][62][63]. Furthermore, the structure of GLI, ZIC3, and Zap1 tCWCH2 domains, in which the two adjacent ZFs form a single globular structure [7,13,47], is clearly different from the structure of the DNA-binding C2H2 ZF domain where each globular ZF unit wraps around the major groove of the DNA [7][8][9][10][11]64] (PDBID: GLI, 2GLI; YY1, 1UBD; WT1, 2RPT; Zif268, 1MEY, 1P47; TFIIIA; 1TF3). In the case of GLI tCWCH2, which is the only tCWCH2 with a known DNA-binding 3D structure, the N-terminal ZF of GLI1 tCWCH2 does not contact DNA [7].
Assuming that the tCWCH2 domain has a function other than DNA binding, what roles could tCWCH2 play? Firstly, our domain structure analysis implies that tCWCH2 likely regulates its neighboring domains, such as canonical C2H2 ZF, WD40, and LIL. Zic/Gli/Glis, PacC, Mizf, Aebp2, Zap1/ZafA, Fungl, Clr1, and Fungl-4ZF possess canonical ZF(s) in the C-terminal flanking region of the tCWCH2. In the case of PacC, analyses of CWCH2 motif-containing tryptophan mutants suggests that DNA binding mediated by ZF2 and ZF3 may be influenced by tCWCH2 motifs in ZF1 and ZF2 [32]. It is possible that tCWCH2 motifs in the other members of this group act similarly in the regulation of neighboring canonical C2H2 domains. In transcription factors, DNA binding domains such as the canonical C2H2 ZF, can define a target sequence and fix the protein position (angle, distance, direction) with respect to the DNA. The tCWCH2 motif could provide a 'hinge' or modulatory junction that controls the relative positioning of other functional domains. Secondly, tCWCH2 motifs may provide a capping structure that isolates the structural influence of N-terminal flanking region. In the case of ZIC3, mutations in its tCWCH2 motif increase the random coil contents more than the zinc-free state [13]. A unique terminal structure in a domain composed of repeating structural motifs is also seen in other conserved domains [65][66][67]. In the case of small leucine-rich repeats and ankyrin repeats, N-capping structures are responsible for protein stability [66,67]. Thirdly, tCWCH2 binding to other molecules is possible, and is observed for many conserved domains. Transcriptional regulator and chromatin remodeling proteins form complexes with many different proteins in order to be functionally active [68,69]. Therefore, the tCWCH2 ZF may not be associated with DNA binding but may serve in protein recognition similar to many C2H2 ZFs [5].
The distribution and conservation of the tCWCH2 sequence motif in various gene classes suggests that this motif has important biological roles. The genes we identified containing one or more tCWCH2 motifs are known to be involved in various biological processes, including chromatin remodeling (Arid2/Rsc9), zinc homeostasis (Zap1/ZafA), pH sensing (PacC), cell cycle regulation, and transcriptional regulation (Zic/Gli/Glis, PacC, Mizf, Aebp2, Zap1/ZafA). In addition, given that several gene classes remain to be functionally characterized (including Fungl, Zfp106, Twincl, Clr1, Fungl-4ZF, Monosiga gene, DdHp), the known functional repertoire of the tCWCH2 motif is likely to increase. We propose that the tCWCH2 sequence motif is a widespread and functional protein structural motif and that further structural and functional analyses are required to fully understand its roles in biological processes.
(4) The tCWCH2 motif contains three additional structural features: (i) bias to hydrophobic residues in the 1 and 2 positions, (ii) longer linkers between the two C2H2 motifs, and (iii) insertion of extra sequences. The structural features suggest that the tCWCH2 motif is a specialized motif involved in inter-zinc finger interactions.

Protein structural analysis
Protein structures were obtained from the NCBI (National Center for Biotechnology Information) database as PDB format files and we used iMol http://www.pirx. com/iMol/ and Swiss-PdbViewer http://spdbv.vital-it.ch/ as the PDB viewer applications. The latter was also used for superimposing multiple protein structures. Magic fit options were selected for the superimposition of structure data because the number of amino acid residues was different between proteins. Zif268 (PDBID: 1P47) and TFIIIA (PDBID: 2J7J) were used for the reference structure of the classical C2H2 ZF. For the distance analysis of the neighboring amino acid residues, we used "Neighbors of Selected Residue Side chain" option of the Swiss-PdbViewer. The software showed the list of the neighboring residues with distances. We dealt with the residues if the same residues were detected as the neighbors in more than 10 models of ZIC3 and Zap1 PDB files that include 20 models respectively.

Database search
We conducted web-based BLAST searches using Pattern Hit Initiated (PHI)-BLAST of NCBI databases http://blast. ncbi.nlm.nih.gov/Blast.cgi. For the sequence pattern of the PHI-BLAST searches, we used "xCxWx(2,35)Cx(5,15)Hx (2,5)Hx(5,25)CxWx(1,3)Cx(5,17)Hx(2,5)Hx". The initial sequence pattern was deduced from the compilation of the Zic family sequences [15,70] and subsequently the intervening sequence lengths were modified to the above pattern until no additional sequences appeared in the search. The seed sequences obtained by PHI-BLAST search were first classified into gene families based on their annotation and on phylogenetic tree analyses (see below). Then a TBLASTN search was performed against (1) all of the following databases: GenBank+RefSeq Nucleotides+EMBL+DDBJ+PDB excluding whole-genome shotgun sequences and expressed sequence tags (nr option in the NCBI, TBLASTN search), and (2) whole genome shotgun sequences of the organisms shown in Table 2. The whole genome sequence TBLASTN search was performed using representative sequences for each gene family (Table 3) initially on a eukaryote-wide basis, and then on a metazoa-and-fungus-wide basis. For the metazoa-and-fungus-wide search, sequences closely related to the selected genes were searched in the following genespecies combinations: Arid2/Rsc9, PacC, Zap1/ZafA, Fungl, Twincl, Clr1, and Fungl-4ZF were searched in fungi; and Zic/Gli/Glis, Arid2/Rsc9, Mizf, Aebp2, and Zfp106 were searched in metazoa. The target organisms are listed in Table 2. These organisms were selected on the basis of the following criteria: (1) high genome sequence coverage (>5×), (2) member of a major Eukaryote taxon or considered to be in a phylogenetically important position. The collected sequences were manually checked for the presence of the sequence of the tCWCH2 motif. In addition, we performed protein BLAST searches using representative sequences for 16 gene families (Aebp2, Arid2, Clr1, Csr1, Fungl, Fungl-4ZF, Gli, Glis, Mizf, PacC, Rsc9, Twincl, ZafA, Zap1, Zfp106, Zic) irrespective of the presence of the tCWCH2 motif. These searches revealed the extent of tCWCH2 motif conservation in each gene family. Redundant sequences derived from a single gene in a single species were removed from the collected sequences except for one representative sequence. We obtained 587 non-redundant sequences containing the tCWCH2 sequence motif.
For the re-evaluation of the initial PHI-BLAST pattern, we generated a tCWCH2 consensus sequence of "LVCKWDGCSEKLFDSPEELVDHVCEDHVGTQLEY-TCLWKGCDRFPFKSRYKLIRHIRSHTGEKP". This sequence was constructed using the following steps: (1) alignment of tCWCH2 by ClustalX software [71], (2) elimination of positions where the frequency of the inserted space "-" was >50%, (3) selection of the amino acid residues most frequently appearing in each position. We performed PHI-BLAST again using four patterns (Additional file 3).
Conserved protein domains in the collected sequences were identified by searching the Conserved Domain Database http://www.ncbi.nlm.nih.gov/Structure/cdd/ cdd.shtml that included domains imported from Pfam and SMART databases.

Sequence analysis
Amino acid sequences of the tCWCH2 motif and one C2H2 ZF in the carboxy flanking were aligned using ClustalX software [71] [76]} were used. To generate the consensus sequence figure, the aligned sequence was subjected to a web-based analysis, WebLogo http:// weblogo.berkeley.edu/ [77]. The PDOC00028 alignment in the Prosite database (12328 sequences, http://au. expasy.org/prosite/) was used for the reference sequence representing the general C2H2 domain. This alignment was entered directly into WebLogo and subjected to linker length analysis. We aligned all of the sequences and then extracted the alignment of each taxon to generate the hierarchically ordered alignment of the consensus sequences. The linker sequence length between the two C2H2 ZF motifs was defined by the number of amino acid residues between the C-terminal histidine residue of the first ZF and the N-terminal cysteine residue of the second ZF.
The PDOC00028 sequence alignment contains the gene name and species together with the start and end position number (e.g., GLI1_HUMAN/301-330). We calculated the linker length by subtracting the end-position number from the start-position number for two ZFs. We excluded linker lengths that were too long (>41), because the average length of general C2H2 sequences (PDOC00028) was 27.95 (n = 12328, standard deviation = 1.39, max. = 42, min. = 10) and the maximum linker length of tCWCH2 was 37.